UCSD Professor Teaches Computers To Listen To Music
UCSD Electrical Engineering Professor Gert Lanckriet is teaching computers to know what you want to hear based on where you are and what you are doing. Going jogging? Your personal music device will play something upbeat. Chilling out after work? It will play something mellow.
Part of the process is teaching computers to classify music. “If you listen to a piece of music and you say ‘this is a romantic song', could we have a computer listen to that piece of music and come to the same conclusion?” he wondered.
So far, the results are promising.
Lanckriet and other researchers are gathering data about music classification with a Facebook game called, "Herd It." Users log on and listen to part of a song. Then, they answer an esoteric question about the music. For example: "What color is the song?" The information is used to teach the computer to recognize patterns in the songs.
"The way to teach a computer what a romantic song is, for example, by showing it 100 romantic songs, and have the computer listen to them," Lanckriet said. "The computer will figure out what these songs have in common and then when it detects any of these typicalities of romantic songs in new songs it has never heard before, it can say, 'this is a romantic song.' "
The technology will also be useful in recommending music to listeners.
Lanckriet said he enjoys playing the drums and keyboards and is happy that his research ties in with his love of music.
“Music is one of these things that really brings people together and for me that’s one of the things that’s personally really important."
Lanckriet was recently named one of the top 35 young innovators in the world by MIT’s "Technology Review" magazine.